0 DOCUMENT RESUME

ED 114 877. CS 501 163

AUTHOR Berger, Charles R. TITLE The Covering Law Model in CoLemunication PUB DATE 75 NOTE 31p..; Paper prese.nted at the 4nnual Meeting of the Speech Commlinicati3On Association '(61St, Houstoh, December'7-30, 1975)

EDRS PRICE MF-50.76.HC-$1.95.Plus Postage DESCRIPTOPS Behavior Theories; *Communication (Thought Tfansfer);' Educational Research; Educational Theories; *Models; *Speech; *TheOrieS IDENTIFIERS Communication Theories; *Covering Law Modelt I ABSTRACT The first Section of thii paper defines -covering law explanation as a theory which maintains that explanation may be achieved, and'may be achieved, by subsuming what is to beexplained under a general law. The modelis examined in light of the deductive-nomological explanation, the deductive-statistical explanation, and the inductive-st'atistical explanation. The second section of the paper examines the'feasibility of.covering law models and examines some criticism of the covering-law approach,'including Paws versus rules, cultural variation, and temporal,variation. Thee third section of the paper-examines potential barriers to theory construction and discusses irrelevant variety,, irrelevant change, realism versus nominalism, 'insensitivity to,embeddedness, and overemphasis on statistical techniques. It is concluded that, a ' defense Oof the covering law approach does not deny the_usefulness of the rule governed approach or the systems approach, and that the covering law approach probably does provide the most complete explanation of a phenomenon when the explanans are,true.- (TS)

I

******************************************************************'*** DOcuments acquired.by,ERIC include many informal' unpublished * * materials.not available from other sources; ERIC makes every effort * ,* to obtain the best- copy available. Nevrtheles's; items of marginal * * reproddcibility, are ften encountered and this .affects the quality .* * of the,microfiche in# hardcopy reproductions EPIC makes available * pia the ERIC Docment ReprodqdtiOn Service (EDRS) .EDRS' is not 4 responsible for the quality of the original document. Reproductions * supplied by EDRS are the best that can 'be made from the original. 4 *****************14*******************)y******************************** U S OEPARTmENTOFHELi," EOOCATION t WELFARE NATIONAL 4NSTITLITEOF EOLICATiOre, : eFE% t,k^ 0, (EC 4E-E .E:, wFPE PSC'. Z.:. PC ,,Y.Fe. OPS ;;;S, l'ATED D'..) ..E:Es- L.e k,,,E sEyr oc; E0,.,CA' 04 0k P7 V

THE- COVERING LAW MODEL IN COMMUNICATION INQUIRY*

*By Charles R. Berger

I

Department of Communication Studies. Northwestern University 1822 Sheridan Road Evanston, Illinois '60201

I .. . 4 1j ` e. fi; , Irl *The author would like to e?tpresshis deep appreciation to Paul H. Arntson, Glen W..Clatterbuck and Gerald R.Miller for their helpful criticisms of an .\ 0 earlic4T draft\ of the present paper.

.10 ,,: * , r;. .i .... - 4. V ry. NNW J THE COVERING LAW MODEL IN CMIUNICATION INQUIRY

Charles R. Berger

NorthWestern University

During tliT 1950's and 1960's,, persons interested in behavioral scientific

approaches to the study of human communication concentrated most of their efforts

on the task of acquiring and tiring various methodological and statistical tech-:

niques. In part, this was a healthy trend at the tithe; however, it turned many

_good researchers in our field into well trained hypothesis testers for theory

builders' in, such disciplines as social psychology, sociology, psycholinquistics,

sociolinguistics and political science. The role models for communication

researchers in terms of'theory construction were found in these and related

disciplinds.

In the past few years, there has be an increase'in the number of papers,

. ji-/' journal articles, and bookS concerned with either issues related to the process

of theory, construction in gen J/,or the presentation of substantive theories

.., or models withincon nunicati j #Fger and Calabrese,'1975; Cappella, 1974;

, / '111/ Cushman and Whiting, lani 41975;.14onge, 1973.; Parks, 1974; andPerce, /1 ! a , 1973). It is probably fa do say that only thrOUgh the development of a unique . 10 ,... 7 / body of theory will commpidation achieve a disciplinary status equal to the . 1,54.-N,'' . . , . . ..1,1 ,. statuses of the other beha ioral sciences. Obviously, methodology and statisticalt .

. . r.7 ... ..$ techniques do not define a discipline. Biology and. sociology employed path

, 1 analysis long before it'wap, .introduced to communication (Dqncan, 1966,and

Wright,..1934, 1954). The employment of. ,arious techniques in communication

, research, no matter how eSoteric the methodology, will not serve to define us

as a unique discipline.

If the primary ta4Lof prese'nt day communication researchers is to develop

and test communicatidnAeeries, it is imperative than they know what theory 41000!". V'

-2-

generating options 5-Fe available to them and what these various options have to

offer. The alternative discussed in the present paper is one which has served

as the dothinant model in both the physical and social sciences: the covering - law model. In the first section of the paper, the basic structure of the cover- .

ing law model will be sketched. We will then examine some criticisms of the

covering law approach. Finally, we will look at some issues surrounding the

realization of covering law theories in communication inquiry. a.

-THE COVERING LAW MODEL

The basic aim of scientific theory is to provide explanations for observed

phenomena. Some might argue that theories also aim to predict; however, explana-

tion tak7sprecedence over prediction., If a theory can provide a complete

explanation for a phenomenon, it can predict the phenomenon. By contrast, it

is possible to have accufate prediction without explanation. In the case of

scientific explanation, the paramount question for which the theoretical structure ti must provide an answer is the 'question."why?". Why do certain groups of persons

have higher rates Of interpersonal conflict than others.?Why did a particular

kind of message induce a large amount of attitude change? Why do persons

engage in more self-disclosure in certain circumstances? These are some questions

.cornunication researchers might wish to answer. The only way.that they can

provide an answer for them is to build a theory. For we might -be able to

predict and cli,:,monstrate with 100% success the rule that the person who answers

the tel;phone always says "hello".before the caller says "hello", but after

observing the phenomenon 1,000 times, we would still not be able to answer yr the question "Why,is it so?". Uncovering a regularity of behavior either in a

"natural-setting"or in a laboratory does not explain it. This issue will .be

explored in greater depth later in the present paper. , -a-

The term "covering law model" was introduced by Dray (1957) in his dis- .

cussion of historical explanation. Dray defined the notion of covering law

explanation as follows:

To put it in a summary way, what the theory maintains is that explanation is achieved, and only achieved,jaysubsuming what is to be explained under a Reneral law. Such an account of basic structure bf explanation is sometimes'referred toas. 40* 'the regularity analysis'; but because it makes use'of the notion of bringing case under a law, i.e. (covering'-it with a law I shall often speak of it hereafter as 'the covering law model.'" (p. 1).

.. Dray cites Braithwaite (1953) and Gardiner (1952) as sources for his definition-

of the covering law model of explanation. For a number of reasons which will

not, be dealt with in the present paper, Dray rejects the position that covering

'law explanations are possible in historical inquiry. He argues that such explana-

tions are likely td be misleading when One attempts .to employ them in,historical 9 research.

A more detailed explication bf the covering law approach tips presented by

Hempel (1965). Hempel distinguished among the following three kinds'of explana-

tions: DeductivelNomological (D-N), Deductive Statistical (D-S), and Inductive '1 Statistical (I-S). Of the three, the D-N explanation is considered to bethe pure case of the covering law model. However, Hempel asserts that both and I-S forms of explanation contain elements ofthepure covering law model.

H considers both D-S and I-S paradigms of explanation to be at leastpartially

subvumed under the general of covering law model. In the remainder of ' this section we will focus our attention on the characteristics of D-N explana- tion and show how it differs from the other two types of explanations. r.

Deductive-Numological Explanation

A 11-N ex planation is divided 'ntothe'following two general parts: (1)

that Which is to be explainedor tie explanandum, and (2) that which does.the;

expi fining, the explanans. The .exp1an.aii3s, in turn, is, dividedinto-two groups

. of tatements. The first type of statement contains particularfacts or initial

o conditions under which the explanandum-phOnomenonwill occur. The second kind

statement expres.ses a trtfl)ormity by means of general law. Taken together,

t e initial conditions and the generallaws give rise` to-the explanandum. In

other words, the explanandumcan he deduced fr6M the combination of the initial

conditions and the general laws. Most theories contain a number ofinitial

'conditions as well as general laws. It is also assumed that the explanandum-

sentence itself is true. Finally, when we refer to the fact that theexplanandum.

is deduced from the explanans,we do not mean Co imply' that only syllogifstic

reasoning is employed in such deductiveoperations. In fact, most theories in

--the physicalsciences use deductivesystems which are far more complex than the

syllogiitic form of deduction presentedin eleAtary logic books.

When we speak of laws in thecase of D-N explanation, we are referring to .j general laws of the form 'All X is Y." Laws:of this type are also'icnownas universal or deterministic.laWs. In the pure D-N system, it is assumed thatthe

laws are true and not restricted by timeor space'(Nagel, 1963). That is, the

law is assumed to hive held-in thepast, to hold in the present, and will hold

in the future.When we assert that every bbdy thatis near the earth ar1d that

freely,fall toward it fall with an acceleration of 32 feetper second per

second,*we are advancing a general law whichapplies to all bodies falling toward

the earth which are near the earth. However, while we may make the claim that

such S law is "universal," it is probablythe'case :that no law in any of the

sciences meets the strict logical criterionof universality. 0 In commenting on

the "universal!' character of laws Kaplan (1.964)asserts: -5-

We might say instead, however, that a scientific law does not have thesform of a strict universal, that this is only an iuealized reconstruction. A unii%ersal proposition, such as might occur in pure mathenuiric, is indeed falsified by even one contrary instance. But in empirical science laws are enunciated and used even though contrary instances are known to be possible, or indeed, when some contrary instances are even known to occur (13. 96)'.

Not only are all laws likely to have exceptions but since it is impossible to

test a law in all possible, circumstances and at all possible-times, it is not

possible to erase all doubt about the possible'truth or falsity of a law.

Hempel (1965) has suggested that when there is doubt about the truth of a

law, the term "lawlike" ought to be uses. Braithwaite .(1953) has made the

same observations concerning the possible falsity. of laws and has adopted the

term "hypothesis" to show the tentative nature of most laws. In practice, then,

D-N theories consist of a set_of_leitial conditions or antecedent conditions,

a set of lawlike. generalizations, and a statement to be explained.

When se assert that a lawlike statement is "general," we do not necessarily

mean that it applies to all possible instances of a given phenomenon. Theories

have both explicit and implicit boundary conditions (Dubin, 1969). The'se

conditions specify the domain of events the theoretical system is designed to

explain. We might construct a theory, containing several general, lawlike state- .. ments, which is only meant to_explain a:phenomenon as it occurs among males or

Americans. For purposes of evaluating the theory,we must confine ourselves to

the intended domain. Moreover, if the initial conditions of the theory are not

met in a particular test of it the data adduced. in such a study are of little

use in evaluating the theory.Frequently, boundary conditions are explicitly

stated by the theory builder; however, boundary conditions afso may be determined

through tests of the theory.

1 A simple example of a D-N system from Braithwaite (1953) might help to

clarify the above,d1scussion. Galileo's law concerning free falling bodies

and thetlower Level hypotheses stlpsumed under it represent a deductive system -6--

ti with hypotheses at three love s f generality:The hypotheses are listed from most to least general as follows:. Z.4

Hypothesis I: Every body near the earth freely falling towards the "earth falls with an accelerationpt 32 feet per second squared. ..(By procedures of integral calculus the .next' hypothesis follows.)

Hypothesis II: Every body starting from rest and freelycfalling towards earth falls 16t` feet in t seconds,whatever number t may be. (Employing the apaicative principle the next hypothesis follows.)

Hypothesis III: Every body starting from rest and freely falling for 1 second towards the earth falls a distance of 16 feet.

We could continue to derive specific hypotheses concerning the relationship between elapsed time and the numbee of feet fallen during that time. For example, = we,would predict that every body starting from rest and freely falling for a period of 2 seconds towards the earth' will fall a distance of 64 feet. We could them derive a.distance prediction for 3 seconds and so on.All of these hypotheses would be at the same level of specificity and would constitute the lowest level hypotheses of the theory. Moreover, these lower order hypotheses would bethe ones we would actually test either theough observationOr some-kind of experimcntal procedure.

There are several characteristics of the above D-N explanatory system that should be noted. First, in the highest level law there is an explicit statement of a boundary condition; namely,. that Only'true "near the earth."

Second, the lowest level hypot,hesis necessarily follows from the higher level ones; that is, the event must occur,within the bou,94ary conditions of the theory, because it follows logically from the premises. Third, if the lowest order hypothesis in the system is not confirmed by experiment or observation, the entire theoretical system as it stands is disproved. However, this is not always the case with a deductive system. Braithwaite (1953 points out that if a lower

of the same order hypothesis followsfrom two or more higher level hypotheses

ES 4

-1;

t, I . .., , .. level, refuatin of the lower level hypothesis will leadto refutation of the ,- t,

-"6onjunctIon of the two or-more higher level hypotheses. At least one of the . .0' , higher'level hypotheses will be shown false by refutation of the lowerorder

*-- - hypothesiS, Thus, in-the case of multiple higher order laws of thesame

level, complete refutationis'not poSsible. Braithwaite (1953) goes on to

assert that in deductive systems.with multiple higherorder hypotheses at

. tIlik.same level, itis possible -to maintain any one hypothesis in the face of

contrary evidence; however, this can only be done by modifyingother hypotheses

in-the sy'stem., A f6urth point worth, noting in our'exampleconcerns confirmation

of the lowest order hypothesis. In the example, confirmation of Hypothesis III

. would not provide complete support for the higher level hypotheses. If we

were to continue to derive lower order hypotheses for differgaA.timeper4ods

aitd each hypothesis were confirmed throughobservation and/or experiment, the

theory would become more fully, supported. However, at no time would we ever

be,ableto assert with complete confidence that the theory is "true"or "proved."

This is the ,base since there will always.be unexamined instances of thehypothesis.

Complete confirmation of theOries is not possible.

What typically happens to the kind of deductive nomologicalsystem examined

in our example is that it is eventually subsumed .bya higher order theoretical

system. Brodbeck (1968), as wellas others, points out that-the higher order

hypotheses or axioms of Galileo's theory became the lowerorder derived hypotheses

or theorems of Newton's theory. Newton's theory, in turn,4'was eventually subsumed

by Einstein's theory. Successive subsumption of one theory,by another indicates

one great strength of the D-N model. It is cumulative and leads to-a systematic

search for progressively more general laws With which to explaina larger number

of empirical events. There are few examples of such successive subsumption

4

0 444 -8-

of theories in the behavioral sciences in general and there are no examples that

I am aware of in the di cipilme of communication in particular. Perhaps suc-

.cessive subsumption of theorieqof_more general theories is the hallmark of a

"mature" discipline: ,

Deductive-Statistical Explanation

In both the physical and social sciences there are lawlike generalizations'

which are,not deterministic in character. Instead, these generalizations are

of a statistical nature; ,that is, the probability that an event X is also of the

type Y is r, or that:

P(

A statistical law asserts that in the long run the proportion of instances of

the type X that are also of the type Y is approximately r. Both the laws of

genetics and the laws of radioactive decay are statistical in nature. We can

predict with a certain probability that if males and females with certain eye

colors have large numbers of children, a'certairi proportion of those children will ,

have a certain eye color. However, we are not in a position to, predict what the

eye color of aparticulat child will be. As BrodJpeek (1968) Has pointed out, . , statistical laws only allowus to predict particular collections of events; . while deterministic laws permit us to make predictions of particular events.

By contrast to a system of explanation, the deductive-statistical

(D-S) mode-of explanhtion 'contains at least one statistical law as part of its

explanans. As the name of the explanatory sy'stem suggests, the D-S explanation . Is deductive in nature; however, because it contains at least one statistical

law, the explanandum statement must include some kind of probability statement.

In a D-S explanation, statistical probabilities which appear in the explanandum

are.derived from probabilities specified in the explanans through the Mathematical,

theory of statistical probability." r'e

-9-

Inductive- Statistical Explanation

A final type of explanatory system explicated by Hemp :l (1965) is that

of'inductive statistical or I-S explanation. In this crse an rt is made to

explain a particular event by employing statistical laws in the explanans.

The followipg example illustrates this situation. Let us assume that we

can make the statement that perceived similarity of attitudes between persons

has a high probability of peodusing a high level of attraction between the two

persons. Let us further assume that Dick and Jane have met and that.they.have

similar attitudes, From this information we,might conclude that it is very

likely that Dick and Jane will bellighly attracted to each other. In this

example, the explanandum statement that,it is very likely that Dick and Jane

will be highly attr acted to each other does not follow with deductive certainty

from the statistical law that perceived similarity leads to interpersonal

attraction with a high probability and the statement that Didk.and Jahe aKe

similar to each other.

Since it is not possible to deduce the explanandum fKom the explanans

with dedUctive certainty in the above example, we must construe the explanans

as inductive support for, the explanandum. To the extent that inductive support

for the explanandum is strong, we can link the explanans with the explanandum

with progressively stronger statements of certainty. However, there are

additional problems with I-S.explanations. For example, in explaining a par-

, ticular event, it is possible to develop explanans with true premises which

yield contradictory conclusions. Let us say that I encounter a 'person, with

a hot temper (T) who also has red hair (R). Upon consulting a social psychologist,

I am told that it is not surprising that I should observe that such aperson

o has a violent temper since the probAbiliCy that a person with a violent temper

will be redheaded is .90. Schematically we might represent this argument as -10- ,0

0 . foliow4; where T stands for a particular person with a ,hot temper.

p(T,R)= .90 I R (.90) p In this argument, our statistical law states that the probability of observing

both hot temper andredheadeduess is .90. We observed a person with red hair (R )

and therefore conclude with 90% ertain67 thit the 'person has a violent temper (T ). AO* Let us assume that if.a person receivesloviritcare from a significant' other

(L), the probability Ghat the person will not display a violent temper (T) is

.85. Further assume that our redheaded friend did in fact receive considerable

loving care from a significant other just prior to our meeting him. The follow- ,

ihg argument could be developed employing this information:

P6,0= .85' VIM

L

(.85), T 13 This argument states that there is a .85 chance that loving care will prgduce

non-hot temper behavior.'Since, our redheaded friend received loving care, there

is a .85 probability that he will not display a violent temper.

The premises of both example arguments are true but they lead to contrary

conclusions about displays ,of temper in the individual case. It is not relevant

that in the second case we were unaware that theviterson in question had received

loving care just prior to our encounter with him. The argument itself is true

and produces a contrary conclusion. Hempel (1965) reTers to the above situation

I o as "the ambiguity of inductive-statistical explanatf . 394) The purpose

of including an example of this ambiguity in the preent discussion was to alert I *

. ... - e ..."... 6 % . . e \ ... , .. *, ,..qh. k ,:* 4 p2. .<, : L_the redder to gt least'onelkof .. .the itficulti!es with I-S,explanatiods. 'HeMpel 5%, discusns a . , numberof ways:"of idea ing with the problem; aswellas other prob- ...... 0

. lems assoaated.vith the I-Si.explanatory system. It should be noted that

. . . the kind of. ambiguity discussed in connection with_the I-Ssystem is. pot e 1 cet . e ' . . . r _,1' possible with a deductive system If the premises of a deauctive,system

tcrue, hen so 'is its conclusion. .A contradictory cbnclusionwhidhis false . ,.. i - cannot be the logicalcoesfquence of premises that'are.equally trul,_ , I' :

0 i *. f 4 ' ....? 1 ..- F. THE FEASIBILITY OF COVERT'IG LAW.V.WELS

_ . -The 'covering law model of explanation has been objecfed'to . . on various -% ...... grOL"ds..*In the following disCussfOnwewill-examine some a hese 7 . .objectpns .. , and atteMpt!..to ascertain their' veracity: Not all possible objectionsto the ...... , . covering lawap-.....proach will be dealt.Within detail in this discussion. -Ohly P 0 major objections to' the model will-be :examfnedt. 'The iritereAed . reader should...... ' >

t . consult Dray (1957) for aerit'-Liq6e,of the covering law model in historical a , 4 . ' explanation.' Hpmpel (1965) alS6 dears withwith criticisms of the .a.pEiroachfrom

both historians, and scientists.

, .....

Laws versus Rules I ,,. s ,

A number of scholars from such disciplinesas antliroioology, sociology . .-. . . aid social psychology have arguedthat since allor most of human -behavior iarf, . rulegoverned, the covering.laW. model of explanation,, withits erdaasis on

' the devel ent of- tawlike generalizations, should be abandbnedin favor of y a search fr the rules which guide human conduct,(Garfinkel, 1967; Coffman,'

'1959; and Secord, 172: and,Toulmin, 190). If one accepts. the notion that

. human-commun tion,is rule governed rather than law governed, thenthe covering r

. la approach or otherapproach which relies on the formation of lawlike'

, . / , gene alizations is of appropriate fol.' communication inquiry. Harreand Secord . - \....._ (1972) have argued that at least for the 'time being, behavio 1 scientists should ..,. , . . , . -, r, , I/ i0 4

I -12- .,L

stop examining variables and their causal relatiOnships with each other and

begin to study in detail episodes and the rules which guide behavior in these 1 epindes. Given this argument, it' would,seem'that not only would the rules

approach dictate the abandormient of the covering law model; but such a para-

digm would also demand that less general "systems analyses" involving variables

(Nonge, 1973) also be discarded. Thus, contrary to Monge's (1973) suggestion,

ti if one views communication behavior as eula governed and accepts Harre and

Secbrd's view, the kind of Blalockian causal analysis (Blalock, 1969) which

R Monge proposes as an alternative to covering law theory is clearly inappropriate

for studying.ruleoverned_behavior. 4

Fortunately; Barre/- and Secord recognize that ultimately the concept of,

rules does litt'e in, the way of explaining behavior. They state:

Though. the concept of rule can be used to explain hOw an actor comes to know what to do, it still leaves open the 'question as to .why. he chose that rule to guide his conduct,. nor does it. explain why he actually acts on the rule, - 'rather than doing- 'nothing. .In explaining action by reference to rule, it is necessary. to add.some further accounts of wants, needs, or expectations of others, the awareness of which would prompt a man"to action in.accordance with the, rule. Thus instancing the rule answers th tion 'How did he know whet to do?' but not the ques ion, 'Why did he do this tUng then and there?'. .(pp. 181-182) li If person explains a sequence of behavior.by recourse to rules; that is,

"The reason Person A in Sit-dation B behaved in ways X, Y,nd Z was because

of certain rules he followed," the.person offering the explanation tuns the

Atk of t.ommiting the.erro'r of hypostatization.

. Atftagt twoquestions can be raised concerning the laws-rules distinction.

, A ,First, we can ask whether,the covering law model is necessary for explaining.

Ehxspersons choose certain rules to guidg their behavior ircertain situations.

01' The major thrust of the present discussion as well as the view given by Harre

and Secord suggest t)lat some kindof coverin law approach would be necessary

Mt in this case. At, some point one must go, bey nd mere description of "what the 14 a.

-13-

rules are" and became concerned with,whysome rules are selected over others., .1 Moreover, the problem of what kinds of environmental cues o? situational

attributes are responsible for the engagement of certain clas.ses of rules

would also seem to be a covering law question.

A second question concerning covering laws and rules involve,s' the

genesis of rules; that is, what social forces produced the kinds of conventions

and "appropriate" modes of behavior we now obserye? This question would seem

to 'be approachable via the covering law model. Perhaps cultures with certain

characteristics have greeting rules which are systematically different fpm

cultures with a different set of characteristics. Such a statement is an

obvious case of a lawlike generalization involving "vai-iables." In addition

to'the question'of genesis of rules, we mi, lso ask,questions concerning

the Processes through i4hieh rules are transi _td from generation to generation.

,Here again, covering law' style approaches from learning theory and social

, 4 modeling (Bandura, 1969) might be employed to explain such phenomena.

The above discussion' suggests,that while the notion of rules may be

of considerable usefulness at the descriptive level, it is still necessary tor'

develop explanatidns fof various manifestations Of rule g9verned behavior.,

Resort td such notions as meta-rules, meta-meta-rules, meta-meta-meta-rules,

. / etc. does not seem to be a' very parsimonious approach to the problem.Moreover,

even if, such a .progression were developed,'it is doubtful that it would

adequately answer the "why" questions raised earlier.What is needed, then,

is a small number of lawlike gen:Iralizations and a set of initial conditions A, which will provide expldnations for, the kinds of regularities observed under

the rubric of rule governed behaviOr.

It is interesting to.note that soze of the researchers who advocate

. the rule governed approach frequently disavow the usefulness of the experiment

/ as a research methodology _(larreand secord, 1972). Instead, it is argued , that such techniques as Garfinkelling are most efficient for'determining

. . 1 r' -14-

. ,

whether or not a parcicular performance is rule governed. In short,

Cdrfinkelling is the practice of behaving in a way calculated to "break the

rule;" so that the consequences of rule breaking can-be observed.To the

extent that the social fabric of the situation is violated, one can be

certain that a rule has actually been broken.According to Campbell (1975)

Carfinkelling is little Ufferent from a quasi - experiment without random-

ization or control, groups. Carfinkelling is equivalent to all experimental

manipulation involving only one value of, the independent variable!

The above theoretical and methodological considerations point to the

conclusion that perhaps the rule governed approach and the covering law

approach a.re best viewed as complementary rather than Competing. As Harre

and Secord (1972) point out, it is almost cert7in that some facets of social

6 behavior, e.g. biological bases, will be best explained by recourse to, some

kind of covering law or causal analysis. Other.facets of social interaction'

will be dealt with best for the present time by employing a kind of rule

governed approaCh. However, it is important to keep in mind that ultimately

the why question will have to be addressed by those. ho take the rule

governed approach.

L

Cultural Variation

It has been argued by some persons that if it is assumed tht communica on

is culturally bound such that symbolip behavior in,one situation is not prod ctive

of symbolic behavior in similar situations, it is not possible to establish

lows. Thus, given cultural variation, the covering law model of explanati n

is inappropriate (Monge, 1973). Earlier in the present paper, it was pointed

out that theories do specify boundary conditions and/or initial conditiot

6

FY Universal generalizations or lawllke generalizationsare made within the

context of such boutdaries. Monge (1973) paints to several physical laws

which he claims g-tel invariant through time and spaCe andare thus universal.

However,, when ore"Aiamines these' physical laws carefully, hefinds that

they ,do not hold in'all cases._Kaplan (1964) provides the following illuS- .

r tration:

When a /student first_learning thegas laws asks why a gas doesn'Ompletely disappear at a sufficientlylow temperature, thus violating the law of conservation ofmass, it is misleading - to infb*Thim only that 'we can' t get to absolute zero'. It is more apiypopiate to make clear to,him that the gas lawsno longer a,Oply, to the same degree of approximation,in the nei.ghbothaod of absolutezero. For that matter, the law of conservation of mass itself is inapplicable in thecontext of nucldar reactions, or if the velocity ofexpansign of the gas appiOa-Ches that of light. (p.. 96)

Kaplan concludes by:assexting that.all lawsare bounded.

The notion of ge'neralization is obviously relative. Newton's theory

was more general thad, Galileo's. Einstein's theory is more abstract than

Newton's. At what point are we willing to assert thatwe live indeed arrived

at a "general law?" PhiSics has progressed at a rapidrate'evek though its

covering law theories Ontain laws which are,not universals In short, it es

seems\ that if one is w ang to admit Viet theory is culturally bound, then,

the problem raised by If lige is obviated. HoweNier, if one wishes to general- ize across cultures and dos' find cultural variations, it is eminently possible . W to find laws which will ekplain cultural variations. Of course, one problem withcrossculturalresearcAs simply that whatmay appear to be "similar situations" between two or4mo'te culturesmay not be "similar situations". ,. \';. at all. It seems reasonable,iCv conclude that even if symbolic behavior is 1 culturally bound, it is still Aossible .to explain such culturalvariations with general covering laws. Moreover, it is possible for theoretical ' propositions' to be invariant fro?h cultureto culture but the character of constructs employed in the theory to change from culture to culture. For

ti 1 . , . example, the lawlike generalization thatincreases tn source credibilityfe . . .

produce increases in persuasion-may beinvariant across cultures; however,

the attributes that make asource highly credible in had may not be the

same attributes that make for'highsource credibility in Costa Rica. The

TED law holds in both cultures, but therelevant dimension. of the constructs

load differently between cultures. 0

Temporal Variation

In addition to cultural variation, temporalchanges in'the behavior of

a phenomenon might be thought of as presentinga problem to the covering law

model. After'all, we found that the covering law l'assumes that its

laws hold throygh time and space. If a phenomenon shows change through

time, for example atlApvelopmental process" Ofsomekind,2it would seem that .1

the covering law model would not be ableto adequately deal with the phenomenon.

Such genetic explanations (Hempel, 1965)are quite common in both history

and the-behavioral sciences.

.Here again we find that if ore is to adequatelyexplainwhy aldev,e1-7 .. ' & opmental process occurs, sets of likestatements will haveto be deve :oped

. , .which explain wh7 the'phendMenonunder study move.s from one 'system stvte" a ' ) if to'another. For example, let us assume that in bothcross sectional

longitudinal studies, -1n whichwe adequately cont of for cohort, d erences, we find that increas*nk age is strongly relatedt the abili to yerform r on concept attainmentv!tasks. The older the child , the better he or she is able to.perform olpth tasks. In viewof our rlier discuspion, this finding is not, by irWel.f, very interesting becauseit has little, if any, ! explanatory power. 4g03rodbeck (1968) has cogently ,61 ointed out, age is a

"non7variable"'becausep it doesn"t explain. . It ,is no until we begin to try to explain why a child of fihe years ofage can perfo say, twice as well '.4

4 -17-

as a_threeyear old on the average that the finding becomes scientifically 4 . r- tnteresting. Of crucial importance to the present discussionis the fact

that in de/eloping an explanation forthe observed age differences, we will

probably have to generate a set of lawlikegeneralizations or "process laws"

(Bergmann, 1957) which explains the temporalchange. These gneralizations

might involve the development of thenervous system. In any case, until

such explanations are developed, thedata regarding age differences, are

merely descriptive. \ Again, it seems that the covering lawmodel i§ necessary for explaining change through time.

POTENTIAL BARRIERS TO THEORYCONSTRUCTION

-Theory-construction is essentiallya creative process. Brodbeck (196$)

points out that it is up to the theory builderto determine which theoretical

f 'statements will be the higher orderaxioms and which statements will be de-

' duced from them as theorems. Of course, the theory builder,mustdecide on

the basis of previous research, hunchor dream exactly what constructs are

likely to be the most important ones in explaining a given phenomenon. By its very nature,,theory construction is a high risk venture. Since theoreticians

are advancing conceptions of how a sliceof physical or soci 1 reality works, **they run the risk of being wrong. It is understandable, thenxJhy Some thedrists

become ego-defensive.when faced withfindings which contradict their eheories.

Their bebiefs about the nature of theworld have been questioned.. ,q, By contrast to theory building,testing a hypdhesis derived from someone

else's theory is a relativelylow risk activity.rbreover, it seems reasonable

to assert that hypothesis testing is generallya less creative activity than

1,s, theory construction; althoughat times research design can demand oonsiderable ./ treativity. Perhaps the "reasonourliscipline 1, is.heavy with hypothesis'testers

is that theory construction has been perceivedas too risky to undertake: t'

_Hopefully, Our discipline 11§' arriving at the point where t at least some persons

Will be willing to risk being wrong! In addition to the perceived risky.nature of theoryconstruction,

there are a number of other "mind sets"which act to discourage the devel-

opment of covering law -type theories. I have been able- to identify five

such impediments to the development ofcovering law theories, but.I am

certain that there are more. In. orde'r, to construct a covering law. theory,

the theory builder fmist overcome the followinglive obstacles-.

Irrelevant Variety

It is a tradition that in their fi,rsUchapters,authors of introductory

psychology textbooks emphasize the complexnature of human behavior.As we

look around us, we are,prone to agree with theassertions of complexity these

A authors make. We have already considered _the problemof cross cultural

differences and their possible impactson our ability to formulate general laws.

However, it just may be that one reason we perceive humanbehavior to be very,

complex is that we do not understand itvery well. I cap remember clearly

during the late 1940's what a mystery the diseasepolio Was.Almost every

sumrer we were' prevented from swimming in public-pools because-it was, e1U

that poiiowas spread in such environments. Cancer is today's medical mystery.

It seems, however, that once the diseaseprocesses are understood, e.g. polio

is caused by a virus from which personscan be immunized, the perceived level ' a . of complexity of the phenomenon decreases drastically. .

In the behavioral sciences, which are still very yojmg,we nay be the

victims of what I call irrelevant variety. Irrelevant variety is generated

by the presence of attributes in esituation whichhave little to do with

the phenomenon we are studying but whichgive the impres'sion that what we

are studying is "very complex." TO illustrate, letus suppose that we are

interested in studying digestion inhumans andwe observe that persons who 1 live iii North America sit on chairs when they eat, whilepersons who live in

2ti Japan and Korea sit on the floor when they eat. "We have observed a reliable

cultural difference in eati.pg habits. However, this kind of cross cultural

'difference may be totally irrelevant to the role played by bile in digestion.

In all likelihood, bile behaves much the same in both cultures' members.

Thus, the observed cross cultural difference introduces variety which is

irrelevant to the phenomenon we are attempting to explain.

Although persons differ along an incredible number of physical,

psychological and soctatdtmensions, merely because this is so does not mean

that all of these differences will make-a difference in terms of the phenomenon

we are studying. It is doubtful that the number of hairs on one's right arm,

eye color, hair color, or cephalic index have much to do with one's susceptibility

to persuasion. For many of the phenomena we study, it is probably the case

that relatively few variables ultimately can account for most of the action.

We just do not know what those powerful variables -are yet. One reason for

this-state of affairs is that we are fooled, more often than not, by irrelevant

variety. r

Irrelevant Change'

Since publication of Bet'ldis book The Process of Communication (1960),

we have been persuaded that comffiunication is best viewed as a, process,: Smith

'(1972) has pointed out that in a majority of cases, persons who do communication

. research,do not, employ research, designs which capture the process nature of

communication phenomena. Since the process notion implies that both physical

and social.'teality are in a constant state of change, here again it wquld seem E. - - -to be impossible to construct laws which are invariant through time. This

,particular problem was discussed_earlier in the present paper,.

However, as is the case with irrelevant variety, it is probable that not

all change has an impact upon.ttle phenomenon we are studying. That is, in

, itt .-20- , II, ' 0..11.

spite of the fact that the physicaluniverse is in a state of (Mange, certain , , f physical laws still hold. %Furthermore,it p ears.that these laws will continue 1

to hold for some tine to come. The same ca be said for ldwlike generalizations

in the behavioral pc ences. For example, in all probability perceivedattitude

similarity will continue to leadto high levels of interpersonal attraction; V, even though our society is in a constant state ofchange. Not all changes through

time make a difference in whatwe study. Recognizing tnis possibility increases

the likelihood that one will successfullybuild a lawlike generalization. - .

.

Realism versus Nominalism ..--.

When theorists go about the job ofconstructing a theory,,theycan choose

to include in their theories onlyconstructs for which there are empirical indicators r. (real constructs) or theycan include some construxts for which there .x. are not yet

t any operational definitions (nominal constructs). This is not only the case in

the social sciences; The construct of mass in physics isa nominal one. The nominal . . --problem with including/constructs in 'a theoryis simply that the theory becomes

more difficult to test as the number of such constructs increases. Theoretical.

statements involving nominal constructscannot be directly.tested. Of course,

there is always the possibility that what is a nominal construct at'thepresent

time will become measureable atsome later time.

There are some goodreasons for including noThinal constructs ina theoretical system. Dubin (1969) has argued that their inclusionwill, encourage the develop-

ment of new empirical indicators; that is, iftheorists insist on including only

real constructs in their theories theyare not likely to search,for new empirical I-

indicaeors. A second argument advanced by Dubinconcerns the ability of a:theoty

to go beyond the "obvious." Here lie argues that sole relianceon real constructs

'will ultimately ledd.to thestudy of relatively trivial problems. In general, / , then, the inclusiOn of ... some nominal constructs in a theory may have the effect of

2f';',. L.

... -21- making the theory Are heuristically provocative. If the nominal constructions do not lead inquiry in new direction's, however, then their inclusion in the theory must be questionned.Moreover, if the statements involving the nominal constructs appear to be very unlikely, then the constructs must be revised or excluded.

The crucial point to keep in mind is that useful theories can be constructed which contain nominal constructs. Nominal constructs can aid in the process of developing covering law explanations. Theorists should not hesitate to use nominal units in their theories if they can provide a plausible Justification for so doing.

In a real sense, one of the eiost imaginative and creative processes involved in theory construction is the development of a nominal construct.

Insensitivity to Embeddedness

The problem of irrelevant variety discussed above centered around the idea that when we are attempting to ascertain, the relationships between two or , ) more variables, we may be distracted from this effort by a host of situational attributes which are, totally irrelevant to the relationships we are Investigating. --41cA The presence of many such attributes leads us to believe that we are dealing with a "very complex phenomenon." There is a closely rented problem concerning our sensit'ivityto the presence of a re]tionship across a number of different situations. For example, let us assume that we are interested in studying the relationship between leader dominance and, the amount of conflict manifested in groups. Suppose thdt we do a series of studies in school classrooms in which we rate teachers on their levels of dominance and also rate their classrooms in term of the, amount of conflict manifested. Say we find that teachers who are highly 4 dominant have less overt conflict in their classes than teachers who are less dominant.

Now, let us assume that we try to study the relationship between the same two variables in a family situation. Obviously, when we move from the classroom

2 r -22-

into the home situation, we are faced with a different kind of so cial conte:tt.

For example, the history of relationships in the family situation is generally

much longer than the histories of relationships in the school situation, Further-

more, and this is.crucial, the ways in which both dominance and-conflict are

. expressed in the two situations may be extremely different. Confkict may be

more overt in the classroom than in the family.Or, dominance may be. more

subtle in the family than in the classroom.

-Suppose our studies in the family situation fail to find any relatibnship

1,atween dominance levels and conflict. Can we confidently say that the dominance- .

conflict relationship does not hold in the family situation? Perhaps not. It

may be the case that given Other measures of dominance and conflict in the

family studies,.we might have found the same relationship between thetwo variables

as we found in the classroom situation. What is being argued here is that geeral

laws Ma34be manifested differently in different situations. The law is generally

true but the way in which it is embedded across situations is such that it looks

"differemt".to the unknowing observer. For example, in the classroom the teacher

might always primar4y.use verbal means to express dominance; while.in the family

situation, dominance might be expreSsed mostly through nonverbalmeans. The,

important point is that the lawlike generalization concerning the relationship

between dominance,and conflict might be generally true, but because the behavioral

manifestations .of it show variations across situations,we might be fooled into

is believing that there is no generalization here.

The above discussion supports the necessity of "triangulation ofmeasure-

ment processes" discussed by Webb, Campbell, Schwartz and Sechrest (1956). Since

most of the research done in our discipline, as well as other behavioral sciences,

does not employ multiple empirical indicators for the same construct, it could

be that we are not finding lawlike generalizations becausewe are not aJequately

measuring the variables we hope to put into a lawlike relationship with each

other. It is imperative that we begin to require that multiple operations?

24 a -23- O

preferablynonreactive ones, be an integral part, ofour research.-

1 Overemphasis on Statistical Techniques

There are at least two ways in which blind faith instatistical techniques

, can prevent us from constructing useful theorieS.The first fstoblem concerns

the use.of such:techniques as factor analysisas substitutes for thinking. The

second problem is related to the notion thatsomethinghas to wary before it

is of interest to us. While I would be the first toencourage everyone who

does communication research to becomeas- sophisticated as possible in as many'

'different statistical techniquesas he can, it is also important to consider

the connection between all of these techniquesand the theory construction'

process.

As an undergraduate student I had a professor whoonce explained while

describing'factor analysis that you only getout of factor analysis what you

put into it, garbage in-garbage out. Eoreover,-he made it clear that since

this is the case, one cannot claim that the factors resultingfrom a factor

analysis constitute some kind of "discovery." Whenone looks at the histOry of

fact6r analysisf hesees that when investigators have attempted to employ the

technique asa theory testing tool, the results have been somewhat'disappointing. ( Spearman's two factor theory led to, the developmentof factoring methods which,

Maximized'the chances of finding the g-factor;while Thurstone developed ro-

tational schemes which increased the probabilitieSof finding several primary

mental abilities. Given kindof situation, it is obvious that what people

"dicover" through factor analysis dependsheavily upon what brand of factor.

analysis they employ and how they label the factors-they find.! The same may be

..sa0.ftii:Lfievcousins.of'factoranalysis.

.-

21; -24-.

There are those who insist on relying on factor analysis and related

techniques to_ "discover" constructs for camnunication inquiry. The general

..itionale for these investigators seems to be that if one makes up a large

,et of items administers them to a large group of persons, and then factors

-fie items, a number of.useful theoretical constructs will be derived. It is

yf further interest to note that very few investigators who firid these new I :.ctors then go on to employ these new variables in further' research. They

:erely report the factor analysis and leave it at that.More important, how-

. _,ver, is the fact that construct fgrmation should precede the use of factor

,Idlysis*and related techniques. nch techniques should not be,emBlgygd as st901 . ,e, : .1 -- - '41'1,f,,e VI, -titutes for conceptual definitiWand the explication o,f relationships among

t _ . . . t- . ,_Ileoretical constructs. Once theoretical constructs have been formed and their

linkages specified in a theoretical system, one should then become concerned

th measurement deVelopment and'employ factor analysis as a means for assessing_

Ale internal consistency of measures and the possible relativships between and

Imong differentempirical indicators of the,construct.

A second problem with an.(5veremphasis on statistical methodology concerns

, . :he fact that when we observe patterns of communication behavior which are

.-Ivarian, or relatively so, wa tend to discount their importance because no

:ariance'can be accouneed for in such a phenomenon! In this"ase, the a ",,tati6tical model forces us to search for/variable so that these variables tan,

..;_t'accounLed for by other variables. However, it seems just as important,

if not mqre important, to un4erstand whsr persons manifest certain relatively

invariant patterris of commUnlcation.through time and space.It is paradoxical

that orthe one hand we say that human communication is a,complex, interactive,

.etc. process, but when weesee recurring patterns of communication we label

.them trivial and not worthy of study because the patterns do not vary much. 47-

.4.`"`

For.exdmple, we,know that when pefsonslieetfor-the firs.t time, their

\ . c .conversations are dominated by exchange's of biogxaphicand demographic informa- . tion. Some observers might find this phenomenon relatively uninteresting.

Although it is easy to understand .why theistnaive observers might judge acon- _ , e -. , ' . versation'dominated by,biOg'raphic and demographicinformation to be "superficial,"

this is the very recurrent kind ofpattern Thich should. pique the curiosity of

behaviooral scientists. Theyshouldask, ",Why do persons exchange thatkind of

information 1.717n they meet?" 'instead, becausethe phenomenon is relatively . , invariant, be vi_ra1 scientists dismiss itbecause they cannot "correlate"x

with'anything else. This is equivalent to dismissing thelaw of falling bodies because it ks invariant!

CONCLUSION

In this paper we have examined thestructure of the covering law model;

some criticismd of it, And some-ways of thinking whichimpede the development

of Covering law theories. .It should beemphasized that a defense,of thecover-

ing'lawapproach doesnot de'ny the.usefulness of the rules governedapproach or the systems approach. It is Probably safe to say, howeyer,that the covering law model does provide'the most complete explanation of a phenomenon whenthe

explanans are true. Morelver, thecovering law approach encourages the'system-

4 atic development of progressivelymore abstract theories through the subsumptive

Process, This process ceases when there isnothing else to explain (Brodbeck,

1968).

The reader- should also keep in, mindthat the covering law modelas described

, by philosophers oscience doesnot desci-ibe how scientists actually behave

(qempel,' 1965; Kaplan, .1964). The explanatory models describedin the present

'paper are ideals.Anyone who does, science,as opposed to merely-talkingabout doing knows that liberal portions of chewinggum and bailing wire are .

, frequently necessary. to hold both theories andresearch together. This is not -26-

only the case in the behavioral sciences; it is also true ih''ekiie physical sciences. I

Braithwaite (1953) has asserted thatpersons will continue to use theories even 9 when they know that the theories are only partially correct. He points out that long before Einstein advanced his theory of gravitation, itwas known that

Newton's theory could not account for the observed motion of,Mercury'sperihelion.

It was not until Einstein's theory was available that Newton'stheory was dethroned. Thus, the reader shoulknot leave thispaper with the idea that in. order tobe successful communication theorists all thatwe need do is "meet the assumptions" of the coveting law model_or.any other modeof inquiry. There is considerable room for hunch, intuition, creativity anda host of other non- .., scientific sounding activities within the domain of scientific inquiry.Overly rigid adherence to any mode of inquiry is not only intellectuallydangerous, but it also takes the fun out of scientific life.

re

a . Re'erences

BanduraA. Principles of behavior modification. New York: Holt;-. Rinehart, and

Winston, 1969.

Berger, C. R. and Calabrese, R.J. Some explorations in initial interaction and

beyond: Toward a developmental theory of interpersonal communication. Human

Communication Research, 1975,1,, 99-112.

Bergmann, G. Philosophy of science. Madisop:.University of Wisconsin Press, 1957.

Berlo, D. K,_The process of communication. New York: Holt, Rinehart, and

. Winston, 1966.

Blalock, H. Theory construction: From verbal to mathematical.formulations. '

Englewood Cliffs: Prentice-Hall, 1969.

Braithwaite, R. B. Scientific explanation: A study of the function of theory,

probability and law in science.N,London: Cambridge University Press, 1953.

Brodbeck, M. Readings dn the philosophy of the social sciences. New York:

Macmillan; 1968.

Campbell, D. T. Personal communication. August, 1975.

Cappella, J. N. A derivation and test of a cybernetic, cooiientational model of

interpersonal commdnication. Paper presented at4he annual convention of ,,the

International Communication ASsociation, April, 1974, New Orleans.

Cushman, D. and Whiting, G. C.An approach to communication theory; Toward

consensus on rules. Journal of Communication, 1972: 22, 217-238.

0 Dray, W. Laws and explanation in history. London: Oxford University Press, 1957. .

4 2:i 4 Dubin, R. Theoryibuitiling: A practical guide to the construction and testing'

of theoretical models. New York: Free Press, 1969.

can, 0. D. Path analysis: Sociological examples.American Journal of'.

Sociology, 1966, 72, 1-16.

Gardiner, P. L. The nature of historical explanation. London: Oxford.bniversit);

Press, 1952.

' A. Garfinkel, H. Studies in ethnomethodoloy.Englewood-Cliffs: Prentice-Hall, 1967.

Coffman, E: The presentation of self in everyday life..New York: Anchor, 1959.

Harre, R. and Secord, P. F. The explanation of social behaviour. London:

Blackwell; 1972.

Hawes, L. C. Pragmatics of analoguing: Theory and model construction in

communication. Reading, Mass.: Addisow-Wesley, 1975.

Hempel, C. G. Aspects of'scientific explanation: And other essays. in thehilosoph

of science. New York: Free Press, 1965.

Kaplan, A. The conduct of inquiry: Methodology for behavioral science.

a San Francisco: Chandler, 1964.

Monge, P. R. Theory construction in the study of communication: The system paradigm.

Journal of Communication, 1973,-23, 5-16.

4 - Nagel, E. Assumptions in economic theory.'American Eoonomic Review Supplementary,

Volume, 1963, 211-219.

Parks, M.R. Toward an axiomatic theory of relational communication. Paper

presented at the annual meeting of the International Communication Association,

April, 1975, Chicago.

3u - Pearce, W. B.itwasenoualrules in interpersonal communication: A reply to

Cushman and Whiting. Journal of Communication, 1973, 23, 160-168.

. Smith,,D. U. Communication research'tind the ideai ofprocess. Speech Moclographs,

1972, 39, 174-182.

Toulmin, S. Concepts and the explanation Of human behavior. In T. Mischel (ed.)

Human Action. New York: Academic Press, 1969.

Webb, E. J., Campbell, D. T., Schwartz, R. D., and Sechrest, L. Unobtrusive

measures: Nonreactive research in the social sciences. Chicago: Rand

1966. Mb

4 Wright, S. The method of path coefficients. Annals of Mathematical Statistics,

1934, 5, 161 -215.

Wright, S. The inprpreLation of multivaria{e systems. In 0. Kempthorne, et al.

(eds.) Statistics and mathematics in biology, 1954;11-33.